U.S. GODAE: Global Ocean Prediction with the Hybrid Coordinate Ocean Model (HYCOM)

1
Global Ocean Prediction with HYCOM
The NOPP GODAE project which developed and tested
the first global 1/12 HYCOM/NCODA
nowcast/forecast system ended in FY08.  Global
1/12 system currently undergoing operational
trials at NAVOCEANO. Formal OPTEST expected in
FY10. An ONR sponsored follow-on to the NOPP
project started in FY09. Includes collaborations
with FSU and NCEP. The principal goal of the
ONR project is to perform the necessary RD to
prepare for the next generation global system at
higher horizontal resolution (1/25 on the
equator, 3.5 km at mid-latitudes, and 2 km in the
Arctic).
2
Global Ocean Prediction with HYCOM

• New capabilities in the system will include
• sub-daily effects - diurnal cycle and tides,
  including the internal tide (excluded in current
  system)
• CICE integration and coupling, including sea ice
  data assimilation
• 3D variational assimilation
• variational adjoint system for determining
  observation impacts on selected measures of
  forecast error

3
Gulf Stream Sea Surface Height Variability

• Along-track from four satellite altimeters
  (Hurlburt and Hogan, 2008)
• b) 1/12? global model year 3
• c) 1/25? global model year 3

(a)
(c)
(b)
4
Mean Simulated vs. Observed Velocity (top) and
Temperature (bottom) Along the Equator in the
Pacific Ocean
Earlier 1/12? HYCOM
Observations
New 1/25? HYCOM
m/s
?C
Observations from the TAO array (Johnson et al,
2007 Prog Oceanogr.)
5
Global Ocean Prediction with HYCOM
Eight largest tidal constituents in a 32-layer
.08 global HYCOM simulation with NOGAPS
atmospheric forcing
Hourly steric SSH
Hourly non-steric SSH
Internal tidal signature seen in the steric SSH
signal
Propagation of tides is consistent with Robertson
and Ffield (2008)
6
Difference in Steric SSH Over 6 Hours June 30
2006 06Z 00Z
7
ESMF Coupled HYCOM/CICE/NCODA
Bering Sea Ice Thickness
no sea ice assimilation
with sea ice assimilation
Black line represents independent ice edge
analysis performed at the National Ice Center.
8
HYCOM Assimilation NCODA
Raw Obs
Navy Coupled Ocean Data Assimilation operational
at Navy production centers (NAVOCEANO, FNMOC)
Automated QC w/condition flags
SST NOAA (GAC, LAC), METOP (GAC, LAC),
GOES, MSG, AATSR, AMSR-E, Ship/Buoy
Profile Temp/Salt XBT, CTD,
Argo Float, Fixed/Drifting Buoy SSH
Jason-1, Jason-2, ENVISAT, In situ Sea Ice
SSM/I, SSMIS Glider Slocum, Sea-glider, Spray
(T,S,U,V)
Ocean Data QC
Innovations
MVOI/3DVar - analysis of 5 ocean variables
temperature, salinity, geopotential, velocity
(u,v)
3D MVOI (3DVar)
Increments
HYCOM
Forecast Fields Prediction Errors
Adaptive Sampling Guidance
First Guess
QC Data Cut
Incremental Update Cycle
9
Assimilation Sub-regions
10
HYCOM/NCODA Runstream
Valid nowcast time
00Z
00Z
00Z
00Z
00Z
00Z
00Z
00Z
00Z
00Z
-120
-96
-72
-48
-24
0
tau
24
48
72
96

• 24 hour incremental update cycle
• IAU applied over the first 6 hrs of forecast (18Z
  to 00Z)
• 5-day hindcast followed by 4-day forecast
• Approximate run time (using 619 Cray XT5
  processors) 13.4 hrs
• six NCODA analyses, nine HYCOM hindcast/forecast
  days
• timing does not include PIPS sea ice coupling

11
SSH Data Timeliness Jan-Apr, 2007Percent data
received by days before nowcast
Illustrates need for 5-day hindcast to obtain
synoptic altimeter SSH data
12
NCODA 3D-Var Data Assimilation

• Observation Space Formulation
• same analysis equation as MVOI (solving the
  equation is very different)
• new HYCOM analysis variables (layer pressure and
  layer thickness)
• new non-separable formulation of correlation
  functions new
  preprocessing options (clustering, profile layer
  averaging, etc)

based on NAVDAS atmospheric 3D-Var system
developed by Roger Daley and Ed Barker at NRL
Monterey
13
So, why 3D-Var?

• global solution no data selection
• much reduced memory footprint over that of MVOI
  important for global HYCOM 1/25? grid
• greater flexibility for assimilating different
  observation data types
• general framework for using more sophisticated
  background error covariance models
• multi-incremental with an inner and an outer
  loop
• outer loop can assimilate observations
  non-linearly related to the forecast state
• valid approximation to 4DVar when assimilation
  time window is short and dynamics are
  quasi-stationary

14
HYCOM Isopycnal Flow Dependent Correlations
Depth/distance cross section of temperature
(contour interval 1?C) and isopycnal following
correlations (color shaded) Observation is at
the surface in the frame on the left, and at 300
meters in the frame on the right
15
END